Direct synthesis of nanopatterned epitaxial graphene on silicon carbide

This article introduces a straightforward approach for the direct synthesis of transfer-free, nanopatterned epitaxial graphene on silicon carbide on silicon substrates. A catalytic alloy tailored to optimal SiC graphitization is pre-patterned with common lithography and lift-off techniques to form planar graphene structures on top of an unpatterned SiC layer. This method is compatible with both electron-beam lithography and UV-lithography, and graphene gratings down to at least ∼100 nm width/space can be realized at the wafer scale. The minimum pitch is limited by the flow of the metal catalyst during the liquid-phase graphitization process. We expect that the current pitch resolution could be further improved by optimizing the metal deposition method and lift-off process.

The research field of graphene patterning initially emerged from the desire to open up a band-gap (BG) in graphene [16], as it has been theorized to depend on the lattice dimensions of the graphene [17]. Using graphene nanoribbons (GNRs), where a single lateral dimension of graphene is narrowed down to the nanoscale, a BG opening that is inversely proportional to the GNR width was later demonstrated [13].
Graphene patterning techniques to date can generally be split into top-down and bottom-up approaches. Due to the availability of large-area graphene sheets, top-down approaches were the first to be developed [16]. They facilitate the structuring of the graphene by mechanically breaking carbon bonds within its lattice or inducing a chemical reaction. Depending on the dimension requirements, repeatability, and accuracy, several patterning techniques have been developed, such as lithography-based techniques [13,18,19], nanoparticle (NP) assisted etching [20,21], carbon-nanotube unzipping [22,23], direct writing techniques [24][25][26], and more.
Plasma-based dry etching, such as reactive ion etching, is currently the most common top-down approach to pattern graphene [19]. Generally, oxygen (O 2 ) plasma is used, and graphene is etched via the oxidation of carbon atoms [27,28]. Similarly, an ultraviolet (UV)-ozone treatment has been shown to transform graphene to graphene oxide (GO), creating local insulating and conducting areas [19].
Metal hard masks [27], photoresists (PRs), or even nanostructures [18] can be used to cover the graphene that is not to be etched. In addition to the formation of edge defects due to the etching itself, some of these processes, in particular the deposition of PRs, may damage the graphene layer itself and introduce additional defects [29]. Recently, maskless approaches, such as focussed ion beam [24], direct laser writing [25,26], and atomic force microscopy (AFM)-based anodic oxidation lithography [30,31], have been explored. They significantly simplify the patterning process and can reduce the risk of contaminating or damaging the graphene during the deposition of the masking layer. However, they suffer from limited patterning speeds and writing areas.
Ongoing research in graphene chemistry combined with established patterning processes has led to novel chemical patterning techniques. Alternatively to etching the graphene, defined areas of an unpatterned graphene layer are covalently functionalized, changing its properties [16]. Furthermore, localized chemical doping or biasing [32][33][34] can be used to achieve the effect of a patterned structure. For example, localized biasing creates patches within the same layer of graphene that have different conductivities [34].
Bottom-up approaches allow for the direct synthesis of patterned graphene. Such approaches are generally based on substrate-assisted synthesis techniques [35][36][37] and liquidphase synthesis of graphene [38]. Most bottom-up patterning techniques carry the benefit of atomic precision, however, they often lack the possibility of creating sophisticated shapes [16].
Recent work has shown the synthesis of patterned CVD graphene using a pre-patterned Ni catalyst layer on top of a Si substrate [39]. This approach requires a subsequent transfer step of the graphene structures to their final substrate to remove the underlying Ni catalyst. Similarly, the site-selective growth of graphene on dielectric and Si substrates using lithography-patterned amorphous carbon as a solid-state carbon source and Ni catalyst layers was demonstrated [9,36].
The graphene demonstrated low sheet resistance. However, only micrometer-sized structures were fabricated.
Finally, the site-selective growth of epitaxial graphene (EG) on Si substrates via the pre-patterning of a SiC film on Si (where the SiC acts as the solid-state carbon source), in combination with a Ni/Cu catalytic alloy has previously been demonstrated [40], as illustrated in figure S1. Since the sidewalls of the patterned SiC are also graphitized, this approach results in three-dimensional graphene-coated SiC structures. This approach has been extensively used to create large-scale graphene-coated SiC micro and nanostructures for MEMS, electronics, and photonics applications [40][41][42]. Note that the above selective bottom-up patterning is inherently tied to the use of SiC/Si pseudo-substrates to obtain a selective graphene growth only where the SiC is present. Also, there are cases where it would be beneficial to only obtain an atomically thin pattern of EG on an unpatterned SiC substrate or SiC film on Si. This approach is still challenging.
In one approach, an aluminum nitride (AlN) capping layer was used to selectively mask the sublimation of Si species from the SiC substrate to obtain EG [35]. This method was shown to significantly reduce the graphene grain size, which was attributed to volatile Al and N atoms during the graphitization as well as AlN-related processing steps. In another work, ( n 110 ) facets of the SiC are used to grow GNRs [37]. The latter approach, while useful to demonstrate the quasi-ballistic conduction of graphene, is limited to the size and the strip shape of the SiC facets.
In this work, we demonstrate that we can extend the Ni/ Cu catalytic alloy concept to directly obtain planar patterned EG micro and nanostructures of a free-form on unpatterned SiC. This new approach is complementary to the already demonstrated method of pre-patterning the SiC, and we anticipate that this approach could also be extended to SiC wafers.

Materials
In this work, we used commercially available SiC/Si wafers from NOVASiC. They consist of unintentionally n-doped, 500 nm thick, 3C-SiC(100) films heteroepitaxially grown on 527 m m lowly-p-doped Si(100). The wafers were diced into 1 cm × 1 cm coupons before they were processed further.

Direct synthesis of patterned graphene
The basic graphene synthesis mechanism makes use of a Ni/ Cu catalytic alloy-mediated graphitization, as introduced in previous works [6,43]. The metal catalysts (10 nm of Ni and 20 nm of Cu) are sputtered onto the SiC layer and placed in a high-temperature furnace (1100 C,  4 10 Torr 5 <´-) where the metal catalysts undergo a liquid phase. Here, the Ni acts as the main catalyst by reacting with the SiC to form Ni silicides (NiSi x ), which enables the release of carbon. Cu has a twofold benefit as it dilutes and distributes the Ni, with the melting point of Cu being close to the annealing temperature, increasing the graphene uniformity, and acts as a catalyst enabling the precipitation and graphitization of the released carbon [6].
Previously pre-patterning the SiC as the solid-state carbon source to selectively grow the graphene patterns on Si would form three-dimensional structures made of SiC coated with graphene (figure S1). Here, we show that it is alternatively possible to pre-pattern the catalytic alloy to define the areas where the graphene will selectively be grown, forming this time a two-dimensional pattern of EG on the SiC (figure 1).
The processing steps are illustrated in figure 1. Individual samples were initially cleaned via sonication (5 min) in / 50 50 Acetone and isopropyl alcohol (IPA). Next, the samples were rinsed with deionized (DI) water, dried using compressed nitrogen (N 2 ) gas, and placed in a Diener Electronics YOCTO-B plasma cleaner (5 min) for further cleaning, see figure 1(a).
The graphene was patterned by pre-patterning of the metal catalysts used to facilitate the graphitization process. This was achieved using a simple lift-off process. Photoresist (PR) lift-off layers were patterned using masked ultraviolet (UV)-lithography for large-area patterning, and electron-beam lithography (EBL), for nanometer patterning, as outlined in sections 2.2.1 and 2.2.2, respectively, and illustrated in figure 1 Subsequently, the metal catalysts, consisting of Ni (10 nm, 99.95%) and Cu (20 nm, 99.999%), were deposited consecutively using direct-current (DC)-Magnetron-sputtering in a Moorfield Nanotechnology nanoPVD S10A sputtering system and lift-off as outlined in sections 2.   large patterned EG on a SiC/Si substrate. The metal catalysts were pre-patterned using masked UVlithography. EG was grown in the shape of the university logo, as can be seen in the (a) optical microscope image and (b) 2D ( 2700 cm 1 -) peak intensity map of the Raman map (2.5 mm 4 mḿ of 50 80 points). The drop-off of the peak intensity towards the edges of the map is an artifact, due to the laser moving out of focus over the large sampled area. silicides, and to reveal the planar graphene patterns, as shown in figure 1(d).
2.2.1. Masked UV-lithography. Lift-off mask patterning: A Suss MJB4 Mask Aligner was used for the masked UV lithography-based patterning of negative PR MicroChemicals AZ nLOF 2020. The PR was spun coated onto the samples using a Laurell WS-650-23PPB Spin Coater (5000 rpm for 60 s). The samples were then soft-baked on a hot plate (110 C  for 60 s), and exposed using the mask aligner. A postexpose bake (PEB) was performed (110 C  for 60 s) before the PR was developed in an AZ 726 MIF bath (50 s), and the samples were rinsed with DI water.
Lift-off: The lift-off of the metal catalysts was performed by initially placing the samples in an Acetone bath for 30 s, following a sonication for another 30 s in Acetone.

Electron beam lithography.
Lift-off mask patterning: EBL was performed using a Zeiss Supra 55VP highresolution field-emission scanning electron microscope (FESEM) and Raith Elphy Plus pattern processor. Positive ALLRESIST E-Beam Resists AR-P 6200.18 was diluted using Anisole to reduce the solids content from 18% down to 4%. It was then spun-coated onto the sample (4500 rpm for 60 s), and the samples were subsequently soft-baked (170 C  for 3 min). For the exposure, an accelerating voltage of 30 kV was used, and the dosage was set to 140 C cm .
was developed in ALLRESIST AR 600-546 developer (20 s), and ALLRESIST AR 600-60 was used as a stopper (30 s) before the samples were cleaned in an IPA bath (2 min).
Lift-off: The metal catalysts were lift-off by placing the sample in ALLRESIST AR 600-71 remover (70 C  for 2 h).

Raman spectroscopy
Raman spectroscopy was performed using a WITec alpha300 confocal Raman microscope that was operated at room temperature and in back-scattering geometry. A 532 nm argon-ion laser (P 30 mW 0 < ) was used for the excitation in combination with a WITec ultra-high throughput spectrometer (UHTS 300 SMFC VIS) and a Zeiss EC Epiplan-Neofluar 100×/0.90 DIC objective, which resulted in a spot size of ∼300 nm. The system was calibrated with a Si reference sample (∼520 cm 1 -). The general integration time for Raman mapping was 0.1 s while the area and step size varied with the map dimensions.

Scanning electron microscopy
Scanning electron microscopy (SEM) was performed using a Zeiss Supra 55VP high-resolution FESEM. Acceleration voltages were set to 5-15 kV to analyze the samples.

Atomic force microscopy
AFM was performed using a Park SE7 AFM. MikroMasch HQ:NSC16/NO AL AFM tips were mounted and used to perform the topographical analysis.

Near-field-imaging
Near-field imaging was performed using a commercial scattering-type scanning near-field optical microscope (s-SNOM) equipped with a pseudo-heterodyne interferometer (Neaspec). The Mid-IR output was obtained by passing the output light of an optical parametric oscillator (OPO) laser (Stuttgart Instruments) that was powered by a pump laser ( 1035 nm, l = 40 MHz repetition rate, and ∼500 fs pulse width), through a difference frequency generation module [44] and further sharped using a monochromator. A parabolic mirror was used to focus the p-polarized emission onto a metal-coated (Pt/Ir) AFM tip (Arrow-NCPt, Nanoworld) oscillating at Ω 280 kHz with a tapping amplitude of ∼80 nm. The back-scattered light was redirected towards a pseudo-heterodyne interferometer for the collection of both amplitude and phase response, demodulated at the second harmonic 2 W of the cantilever oscillation frequency [45].

Patterning using masked UV-lithography
The selection of the lithography process is crucial for the scalability and maximum obtainable resolution of any patterning approach. The MJB4 Mask Aligner used in this study enables the exposure of 4" wafers. Under optimal conditions, a maximum resolution of 0.5 m m can be achieved.
As a first demonstration, showing the possibility of patterning arbitrarily large and complex graphene patterns, we have transferred the university logo onto the PR masking layer on top of a lowly p-doped 3C-(100) SiC/Si coupon. The dimensions of the structure were ∼2 mm 4 mm, and an optical microscope image of the final patterned graphene in the shape of the university logo can be seen in figure 2. It also shows the Raman spatial map (2.5 mm 4 mḿ of 50 80 points) of the 2D peak (∼2700 cm 1 -) to confirm the graphitization ( figure 2(b)). Subsequently, planar 2 m m wide graphene gratings with a 6 m m pitch were fabricated, exploiting the maximal resolution of the used equipment. Figure S2 shows an optical microscope image of the patterned graphene gratings on the SiC/Si substrate as well as the D (∼1350 cm 1 -), G (1580 cm 1 -), and 2D (∼2700 While the structures were fabricated on 1 cm 1 cḿ SiC/Si coupons for the sake of this demonstration, the approach can be scaled up to create wafer-scale graphene patterns with UV-lithography.

Patterning using electron beam lithography
EBL was used to produce nanometer-sized graphene structures. EBL was performed using a Zeiss Supra 55VP FESEM and Raith Elphy Plus pattern processor. In order to explore the capabilities of our patterning process, the mask consisted of four sets of graphene gratings and circular resonators with widths/diameters of 50 nm, 100 nm, 200 nm, and 400 nm, and their respective pitches being 100 nm, 200 nm, 400 nm, and 800 nm.
Due to the dimensions of the fabricated structures, SEM was used for the optical inspection of the fabricated samples. Figure 3(a) shows an SEM image of the well-resolved patterned metal catalyst after lift-off, while figure 3(b) shows the SEM image of the patterned graphene after the Freckle etch. It should be noted that the contrast of the graphitized areas, in comparison to the SiC layer, depends on the layer count as well as the SEM voltage and working distance (WD) [46]. In general, thicker graphene, i.e. few-layer graphene, appears darker and thinner graphene, i.e. monolayer graphene, lighter in color. To resolve the nanostructures, a relatively high acceleration voltage of 5 kV and low working distance of 4.3 mm were used in this study, which are not ideal conditions for the imaging of graphene.
SEM images of the 50 nm (100 nm pitch) and 100 nm (200 nm pitch) gratings in figure 4, show the presence of graphene (dark regions). Darker regions show the location of thicker EG, while brighter regions within the gratings indicate thinner graphene, which have lower contrast to the SiC [46]. The catalytic alloy-mediated graphitization of SiC has previously been shown to form few-layer graphene with up to 7 layers [47].
Furthermore, the images also indicate some growth of EG in areas beyond the initial catalyst pattern. This leads to potential shorting of the graphene gratings as their pitch decreases. This effect is not pronounced on the 100 nm gratings but obvious for the 50 nm graphene gratings. This suggests that the lower pitch resolution of this patterning approach is roughly equal to or smaller than 200 nm.
We attribute this pitch limitation to the liquid flow of the metal catalyst during the graphitization process at 1100 C  and a residual and undesirable metal build-up due to the sputtering and lift-off process.
The flow is a consequence of most of the catalysts being effectively in a liquid phase during the annealing step [6,43], which results in a loss of line definition, and hence resolution, as illustrated in figure 5(a). The loss of the line definition is hence transferred to the EG layer pattern. This issue becomes a limitation as the structures get smaller, however, it appears less crucial for larger ones, as schematically illustrated in figures 5(b) and (c) exemplifying 50 nm and 200 nm gratings, respectively.
In fact, the AFM profile across a 400 nm metal catalyst grating before the graphitization (figure S3) shows the excess metal build-up of about 15 nm thickness at the edges of the metal pattern. We attribute this build-up to the accumulated metal on the PR sidewalls during the sputtering process, which is not effectively eliminated during the lift-off process. We expect the metal build-up to exacerbate the flow effect and contribute to further loss of line definition. Optimization of the deposition and lithography approach to reduce the side wall deposition, for example, using highly directional evaporation techniques, is hence expected to mitigate the build-up and thus the extent of the catalyst flow during the graphitization, improving overall the pattern definition and resolution.
Raman large-area mapping was conducted on a large graphene patch (located on the same sample as the university logo) and 400 nm graphene structures to further characterize the patterned graphene. The D (∼1350 cm 1 -), G (1580 cm 1 -), and 2D (∼2700 cm 1 -) peak intensity maps of the Raman map  maximum, and minimum of the / I I D G ratio were determined to be 1.13 and 0.18, respectively. Therefore, we can estimate the minimum and maximum of the grain as ∼17 nm and ∼107 nm, respectively. Note, the Raman ratios and estimated grain sizes are in line with large-area graphitized samples.
Furthermore, two Raman spectral maps (5 m 5 m m ḿ of 20 20 points) were performed on the 400 nm graphene resonators and gratings. Figure 6(b) shows the 2D peak intensity maps of the measurements and their locations. The patterns can be clearly seen and distinguished from one another. However, our Raman spot size of ∼300 nm impedes the accurate evaluation of the nanostructures.
Near-field imaging was performed using s-SNOM to analyze the patterned graphene at the nanoscale further. Figure 7 shows near-field imaging measurements of 100 nm and 400 nm gratings that were excited by tuning the laser source at 9.6 m. m The near-field imaging clearly highlights the resonances of individual and groups of graphene grains within the gratings. Figures S4 and S5 show the amplitude and phase plots of the 100 nm and 200 nm gratings, respectively. While the near-field imaging results at a single wavelength may give the impression that some gratings are not fully graphitized, further near-field imaging of the 400 nm gratings using additional wavelengths of 7.5 m, m 8.5 m, m and 10.7 m, m as illustrated in figure 8, show that graphene grains of different sizes resonate at different wavelength excitations. The change in the background intensity for the 10.7 m m amplitude map results from the high density of states related to the excitation of surface phonon polaritons in SiC [49].
Note that our previous work has indicated the support of plasmonic resonances in EG grown on SiC nanowires using the catalytic alloy-mediated graphitization approach used in this work [41].
The near-field images further show that some of the graphene grains conform to the shape of the gratings, particularly for the smaller 100 nm gratings, see figure 7. This confirms that some of the grains approach ∼100 nm sizes.

Conclusion
In this work, we introduce the direct synthesis of planar (2D) micro and nanopatterned epitaxial graphene on silicon carbide that can be carried out at the wafer-scale. This is done by pre-patterning the Ni/Cu catalyst metals using simple lift-off processes prior to the graphene growth. This approach eliminates the need for chemical or ion and laser-based etching of the graphene, which carry the risk of potentially damaging or contaminating the two-dimensional layer. While this demonstration was carried out on a SiC/Si pseudo-substrate, this methodology could be extended to EG on bulk SiC wafers.
We estimate that the resolution of this patterning approach can currently enable grating sizes down to roughly ∼100 nm width/space. This limit is mainly determined by the flow of the metal catalysts which are in liquid phase during the graphitization, causing a loss of definition of the EG pattern. This effect is exacerbated by the metal build-up at the edges of the metal patterns due to the low anisotropy of the sputtering deposition in combination with the lift-off process.
We hence anticipate that the current pitch resolution could be improved by reducing the build-up using a highly directional metal deposition method.
Note also that larger graphene grain sizes using this method may be obtained through a further fine-tuning of the Ni/Cu alloy composition, a stricter control of cooling/heating ramps and furnace pressure, or using a different alloy altogether. Such optimization goes beyond the scope of this work, however, any of those additional improvements could be used in combination with the selective patterning method introduced in this manuscript.

Data availability statement
All data that support the findings of this study are included within the article (and any supplementary files).

Funding
We gratefully acknowledge support from the Australian Research Council through the Center of Excellence in Transformative Meta-Optical Systems (TMOS) CE200100010 and through the Center of Excellence Future Low-Energy Electronics Technologies (CE170100039)